Residual echo suppression

Claim 2

In a hands-free telephone comprising a loudspeaker coupled to an output source for generating an output signal, a microphone for generating an input signal and an acoustic echo cancellation means for generating a reduced-echo signal, a method for further reducing echo comprising the steps of: determining an echo canceller energy ratio (ECER) as an energy ratio between an output of said acoustic echo cancellation means and said output source output signal; comparing ECER to a predetermined threshold level; performing spectrum reshaping of said reduced-echo signal if ECER exceeds said predetermined threshold level.

Claim 3

The method of claim 1 wherein said acoustic echo cancellation means and said spectrum reshaping step are in a sub-band domain.

Claim 4

The method of claim 1 wherein the spectrum reshaping step comprises the steps of generating an output of a noise suppression circuit in said acoustic echo cancellation means; computing an output signal r i (t) as r.sub.i (t)=s.sub.i (t)*NS.sub.i *ξ.sub.i where s i (t) is the output of the noise suppression circuit, NS i is the noise spectrum estimate and ξ i represents the scaling value based on the long term average power of the s i (t).

Claim 5

The method of claim 3 wherein ##EQU6## where λ is a constant.

Claim 6

The method of claim 4 wherein λ is a time-invariant constant determined empirically.

Claim 7

The method of claim 4 wherein ##EQU7## where M is the number of bins and N is the total number of data points in a frame.

Claim 8

The method of claim 3 wherein ##EQU8## where λ(t) is a time-varying constant and Ψ i (t) is a long-term power estimate for the i-th bin.

Claim 9

The method of claim 7 wherein ##EQU9## where M is the number of bins and N is the total number of data points in a frame.

Claim 10

The method of claim 7 wherein Ψ.sub.i (t)=(1-μ)Ψ.sub.i (t-1)+μ|s.sub.i (t)|.sup.2 where μ<<1.

Claim 11

The method of claim 9 wherein ##EQU10## where M is the number of bins and N is the total number of data points in a frame.

Claim 12

In a hands-free telephone comprising a loudspeaker coupled to an output source for generating an output signal, a microphone for generating an input signal and an acoustic echo cancellation means for generating a reduced-echo signal, apparatus for further reducing echo comprising: means for determining an echo canceller energy ratio (ECER) as an energy ratio between an output of said acoustic echo cancellation means and said output source output signal; means for comparing ECER to a predetermined threshold level; means for performing spectrum reshaping of said reduced-echo signal if ECER exceeds said predetermined threshold level.

Claim 13

In a hands-free telephone comprising a loudspeaker coupled to an output source for generating an output signal, a microphone for generating an input signal and an acoustic echo cancellation means for generating a reduced-echo signal, apparatus for further reducing echo comprising: means for determining an echo canceller energy ratio (ECER) as an energy ratio between an output of said acoustic echo cancellation means and said output source output signal; means for comparing ECER to a predetermined threshold level; means for performing spectrum reshaping of said reduced-echo signal if ECER exceeds said predetermined threshold level.

Claim 14

The apparatus of claim 11 wherein said acoustic echo cancellation means and said spectrum reshaping means operate in a sub-band domain.

Claim 15

The apparatus of claim 11 wherein the acoustic echo cancellation means comprises a noise suppression circuit; and the spectrum reshaping means comprises: means for computing an output signal r i (t) as r.sub.i (t)=s.sub.i (t)*NS.sub.i *ξ.sub.i where s i (t) is the output of the noise suppression circuit, NS i is the noise spectrum estimate and ξ i represents the scaling value based on the long term average power of the s i (t).

Claim 16

The apparatus of claim 13 wherein ##EQU11## where λ is a constant.

Claim 17

The apparatus of claim 14 wherein λ is a time-invariant constant determined empirically.

Claim 18

The apparatus of claim 14 wherein ##EQU12## where M is the number of bins and N is the total number of data points in a frame.

Claim 19

The apparatus of claim 13 wherein ##EQU13## where λ(t) is a time-varying constant and Ψ i (t) is a long-term power estimate for the i-th bin.

Claim 20

The apparatus of claim 17 wherein ##EQU14## where M is the number of bins and N is the total number of data points in a frame.

Claim 21

The apparatus of claim 17 wherein Ψ.sub.i (t)=(1-μ)Ψ.sub.i (t-1)+μ|s.sub.i (t)|.sup.2 where μ<<1.

Claim 22

The apparatus of claim 19 wherein ##EQU15## where M is the number of bins and N is the total number of data points in a frame.

Claim 23

In a hands-free telephone comprising a loudspeaker coupled to an output source for generating an output signal, a microphone for generating an input signal and a sub-band acoustic echo cancellation means for generating a reduced-echo signal, a method for further reducing echo comprising the steps of: determining an echo canceller energy ratio (ECER) as an energy ratio between an output of said acoustic echo cancellation means and said output source output signal; comparing ECER to a predetermined threshold level; performing sub-band spectrum reshaping if said reduced-echo signal of ECER exceeds said predetermined threshold level, said reshaping step comprising the steps of: providing a noise suppression circuit in the acoustic echo cancellation means; computing an output signal r i (t) as ##EQU16## where s i (t) is the output of the noise suppression circuit, NS i is the noise spectrum estimate, ##EQU17## where M is the number of bins and N is the total number of data points in a frame; and Ψ.sub.i (t)=(1-μ)Ψ.sub.i (t-1)+μ|s.sub.i (t)|.sup.2 where μ<<1.

Claim 24

In a hands-free telephone comprising a loudspeaker coupled to an output source for generating an output signal, a microphone for generating an input signal and a sub-band acoustic echo cancellation means for generating a reduced-echo signal, a method for further reducing echo comprising the steps of: determining an echo canceller energy ratio (ECER) as an energy ratio between an output of said acoustic echo cancellation means and said output source output signal; comparing ECER to a predetermined threshold level; performing sub-band spectrum reshaping if said reduced-echo signal of ECER exceeds said predetermined threshold level, said reshaping step comprising the steps of: providing a noise suppression circuit in the acoustic echo cancellation means; computing an output signal r i (t) as ##EQU16## where s i (t) is the output of the noise suppression circuit, NS i is the noise spectrum estimate, ##EQU17## where M is the number of bins and N is the total number of data points in a frame; and Ψ.sub.i (t)=(1-μ)Ψ.sub.i (t-1)+μ|s.sub.i (t)|.sup.2 where μ<<1.